23
In the upper sequence the results obtained from individual
operators are arranged with respect to increasing mean errors,
where the numbers above give the number of plates measured by
each operator and the arrows indicate the spread of the mean
errors for each operator. The range of the average per
formance for the 34 operators is from ±1.1 to ±2.2 micron.
The lower diagram presents the sequence of mean errors
for operator #6, who has one of the smaller average mean errors
(±1.37 micron), for a sequence of 78 plates covering an
observation period of 15 months. Aside from showing the
variations in performance, one may wish to discover a trend
of improvement by a slight reduction of the average mean error
with time. The presented curve is typical for the performance
of the other operators, at their individual mean performance
levels.
Figure 10 is a histogram formed from the 645,872 com
puted differences of the corresponding double measurements.
Its adherence to a normal distribution is evidence that the
measuring process as such appears to be sufficiently free of
bias errors*, consequently, the average mean error of ±1.63
microns can be accepted as an indication of the average pre
cision in the measuring operation. The overall range for
precision of measurement for individual plates is from ±0.76
to ±2.4 micron. Both the instantaneous operator performance
and slight variations in image quality, due to seeing con
ditions, are sources affecting the differences in the pre
cision of the measuring process.
b. Analysis of the Accuracies Obtained in the Recon
struction and Orientation of the Photogrammetric
Bundle, Based on 930 Plates Reduced for the World
Net Program Between July 1966 and June 1968
The parameters considered necessary to establish the
projective relationship between the updated star catalog
data (compare Formula (3)) and the measured star image
coordinates were given in Figure 6.
The reference stars are stored on a tape, selected from the
catalog of the. Smithsonian Astrophysical Observatory and
limited to stars < 8th magnitude with a positional accuracy
of < ±0"4. Corresponding weights are applied during the least
squares solution for the single camera. The resultant mean
error for the image coordinates as obtained from the